Guardrails AI is an open-source Python framework that checks what goes into and comes out of an LLM, and helps get structured data back from it.
In plain words
Guardrails wraps your model calls with guards. Each guard is built from validators, small checks for a particular risk, which you can take from a shared collection called Guardrails Hub. Guards inspect both the prompt and the response and can detect and fix problems. The same framework also helps you get output in a fixed structure. In 2026 the project began moving its validators to ordinary pip packages and set 25 August 2026 as the planned end of its hosted remote checking service.
A quick example
Imagine an app that asks a model for a JSON summary of a customer email. A guard checks the reply has the right fields and contains no personal data before the app uses it.
Where you'll meet it
- Validating LLM outputs in Python apps.
- Getting structured data from LLMs.
Related
Sources
- DocsGuardrails README, Guardrails AI (GitHub) · read 29 Sept 2026
- DocsGuardrails repository, Guardrails AI (GitHub) · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.